Summary: This I/O chassis has much lower contamination from fans and power supply magnetic fields than previously tested I/O chassis. Coherence between magnetic fields produced by the power supply and the fans and empty ADC channels was less than 0.001. However, there is slight coherence between empty ADC channels, and I suggest a future test with a linear supply providing the 24V power.

Previous I/O chassis models have had small cooling fan peaks evident in all tested empty ADC channels (https://alog.ligo-wa.caltech.edu/aLOG/index.php?callRep=17944). These features can appear in real data when the signal is near the ADC noise floor (https://alog.ligo-wa.caltech.edu/aLOG/index.php?callRep=11931 , https://alog.ligo-wa.caltech.edu/aLOG/index.php?callRep=2895 ). These investigations showed that the fans coupled by power supply ripple, not magnetic fields.

In addition, in older versions of the I/O chassis, magnetic fields from the on-board power supplies coupled into ADC channels (https://alog.ligo-wa.caltech.edu/aLOG/index.php?callRep=2895 ). A different power supply in the later I/O chassis versions reduced magnetic fields, in some bands by orders of magnitude, to where the power supply magnetic fields were no longer evident in coherence with blank channels, though the fan peaks were still present (https://alog.ligo-wa.caltech.edu/aLOG/index.php?callRep=17944 ).

We tested the latest I/O chassis in the H2 room test stand, using a magnetometer to identify the fan peaks and detect fields emitted by the power supply and other electronics. To identify the fan peaks in the magnetic spectrum, I partially covered the fan ports with paper in order to manipulate the fan rotation frequency (Figure 1). 

Figure 2 shows the magnetic field spectra and the coherence between the magnetometer channel and blank channels. No coherence is evident, at a coherence level of less than 0.001, with either the fan peaks (the broad peaks just to the right of 60 Hz harmonics), or the broad signal from other electronics. The broad band fields from the power supply were roughly three times lower than from even the best previously tested I/O chassis,  (https://alog.ligo-wa.caltech.edu/aLOG/index.php?callRep=17944 ).

Figure 3 shows spectra for all empty ADC channels, showing that fan peaks were not evident in 5000 averages but that there were small features that appeared in all channels, behaving like intermodulation peaks between high frequency signals. These features produce coherence between empty channels, shown in Figure 4. The lack of coherence demonstrated in Figure 2 indicates that this contamination is not due to magnetic field coupling. 

One possibility is that these small features are associated with the 24 V supply to the chassis, which is now from a Sorenson switching supply. I have left the test setup in place so that the 24V supply can be replaced with a linear supply for a future investigation.

Robert, Dave, Jonathan